California plans support for autonomous taxis without backup drivers

The fallout from the fatal crash on March 18th when an Uber self-driving car killed a pedestrian has naturally led rival taxi groups developing autonomous fleets both to exploit the situation commercially and to divert blame from the industry as a whole. It looks like they may be successful on both counts given that Uber has suspended its autonomous driving development program indefinitely, while the industry as a whole received a boost when California’s public utility regulator issued a proposal that would allow self-driving car companies to transport passengers without need for a backup driver in the vehicle. This would obviously greatly reduce costs, which is one of the whole points of autonomous driving for taxi companies, while freeing up a space for an extra passenger.

It could be argued this signal from California, which remains to be ratified in a vote, was issued despite or even because of the Uber case. Analysis so far has suggested all three parties were partly culpable, the autonomous system, the safety driver who looked down at the crucial time and also the victim, who took less care than she should have done. The last point should cut little ice, since cars should be driven with enough margin for error to avoid running somebody over even when the pedestrian is at fault, save for those rare cases when it is unavoidable as when somebody deliberately jumps in front of a vehicle.

It also seems that a human driver paying full attention should have been able to pull up in time, or at least slowed to a speed that might have been non-fatal, with indications that the autonomous system was not functioning as well as it should. It is difficult to assess claims by rivals such as General Motors and Alphabet’s Waymo that their technology would also have averted the accident, but what is certain is that their vehicles too will be involved in a fatal crash sooner or later.

The issue is poignant because autonomous driving advocates claim it will greatly reduce incidence of fatal vehicle crashes, given that on the day of the Uber fatality around 100 others died on US roads, often as a result of human error, recklessness, rage, lack of attention or drunkenness. But autonomous driving will only gain full public support if it is seen to be far safer than the status quo and so a death at this stage is untimely. On average in the US, there is about one fatality for every 86 million vehicle miles traveled and automated vehicles have only run about 20% of that. Of course this is not statistically significant, but until the miles of fatality-free autonomous driving have accumulated way beyond the 100 million-mark doubts will remain.

But the Uber case also questions the value of safety drivers at all with the implication that the whole motor industry would be better jumping straight to SAE Level 4 on the autonomy scale, which is fully autonomous but in controlled areas. Levels 2 and 3, which is where the industry is really at now, requires presence of a driver on standby even if control is handed over at times. The California Public Utilities Commission, which regulates utilities including ride-hailing apps, seems to have brought the argument that the overlap between human and autonomous driving embodied in Level 3 especially brings risks of its own that manifested themselves in the Uber fatality.  Especially over a long journey or taxi session, humans are prone to loss of concentration or even nodding off to sleep if they are not actually doing the driving most of the time.

Uber is now in the ironic position of hoping that regulators will apply the brakes to autonomous driving rollout to provide time for investigations into its crash to be completed and appropriate revisions to the technology and operations to be made. After all Waymo is now well placed as the only company with a fleet of fully self-driving cars on public roads in the US. Furthermore it is already supporting Level 4 operation and is poised to launch the first truly robotic taxi service by the end of 2018 – if regulators allow.

As things stand, Waymo already has permission from the Arizona Department of Transportation to operate its driverless fleet of Chrysler Pacifica minivans to pick up paying passengers in the state, through a smartphone app or website. It has 600 vehicles and more on order, with plans to kick off in Arizona’s largest city Phoenix later in the year. Unless its hubris over the safety of its system comes back to haunt it, Waymo could gain a massive head start in autonomous driving as a result of the Uber crash and this proposed California law. Being able to ‘prove’ the safety of its technologies could be a stunning blow to the rest of the industry.